How Parsio processes your data across its parsing engines

Parsio isn't a single one-size-fits-all pipeline. You choose the parsing engine that fits each mailbox — AI Parser, GPT Parser, Template Parser, or OCR Converter — and each one has a genuinely different data path, with different vendors (or none at all) involved. We want to be precise about what happens in each case, so you don't have to take it on faith.

This page complements our general Security page with a closer look specifically at how each engine handles your data.

Four engines, four different data paths

AI Parser runs your document through pre-trained, prebuilt extraction models — no configuration, no prompt, no template. GPT Parser sends the document's text (or an OCR'd version of it) together with a prompt you write, to a large language model, for documents whose layout and wording vary too much for fixed rules. Template Parser matches your document against a template you build by pointing and clicking on fixed field positions — no AI or LLM vendor is involved at all. OCR Converter extracts raw text and tables from scans, PDFs, and images, without interpreting them into structured fields.

AI Parser — pre-trained models, zero configuration

AI Parser uses Microsoft Azure Document Intelligence's prebuilt models, each specialized for a document type (invoices, receipts, ID documents, business cards, bank statements, and more). You don't write a prompt or build a template — the model already knows what to look for. There's no free-text instruction sent anywhere; only the document itself is processed by the model.

GPT Parser — your prompt, an AI model

GPT Parser is built for documents whose layout or wording is too inconsistent for fixed rules. You write a plain-language prompt describing the fields you want; Parsio extracts the document's native text (or runs OCR first if it's a scan or image) and sends the text and your prompt to OpenAI for extraction. That data is encrypted in transit and is never used to train OpenAI's models.

Template Parser — no AI, no third-party model call

If you'd rather not involve any AI vendor at all, Template Parser is a genuine no-AI path: you define a template by marking field positions on a sample document, and every future document with that layout is matched against it using rule-based logic that runs entirely on Parsio's own infrastructure. No document content is sent to OpenAI, Microsoft, Mistral, or any other AI provider for this engine.

OCR Converter — text and table extraction

OCR Converter turns scans, PDFs, and images into raw text and tables using AWS Textract or Mistral's OCR, depending on configuration. It doesn't interpret the output into structured fields the way the other three engines do — it's a text-extraction step, often used as a building block ahead of Template Parser or your own downstream processing.

Training and data retention

• We never train or fine-tune any model on your data, and we never sell your data. This is a standing commitment, not a plan-dependent feature.
• Our agreements with our AI subprocessors — OpenAI, Microsoft, and Mistral — prohibit them from using data submitted for extraction to train their models.

Encryption and transport

All communication with our AI subprocessors happens over encrypted connections (HTTPS/TLS 1.2+), the same standard we apply to every other part of our infrastructure.

Data segregation

Each Parsio account's documents, templates, and results are logically isolated from every other account. Processing one customer's document never exposes or mixes in another customer's data.

What stays under your control

• You are the Data Controller; Parsio and its subprocessors act only as Processors, on your instructions.
• You can delete any document, template, or your entire account at any time.
• Configurable retention policies let you automatically delete processed data between 1 and 180 days.
• You choose which engine — and therefore which vendors, if any — touch each document, mailbox by mailbox.

Frequently asked questions

Which engine should I use if I don't want any AI vendor involved?

Template Parser. It's rule-based field-position matching that runs entirely on Parsio's own infrastructure — no document content is sent to OpenAI, Microsoft, or Mistral.

Is Parsio just a wrapper around ChatGPT?

No. Only GPT Parser calls OpenAI, and only for documents you route to it. AI Parser uses Microsoft's prebuilt extraction models, Template Parser uses no AI vendor at all, and OCR Converter uses AWS Textract or Mistral for text extraction only.

Does the raw file get sent to AI providers, or just extracted text?

It depends on the engine. GPT Parser sends extracted text (native or OCR'd) plus your prompt to OpenAI. AI Parser and OCR Converter process the document directly through their respective provider's model. Template Parser never sends document content to any AI vendor.

Do you use embeddings or vector search for long documents?

No. Parsio doesn't currently use embeddings or vector retrieval — GPT Parser sends the extracted text and your prompt directly.

Can Parsio employees read my documents?

Access to customer data is restricted by least privilege and limited to what's needed for support or troubleshooting, and is audited regularly. See our Security page for details.

Is my data used to improve Parsio's or its AI providers' models?

No. We don't train on your data, and our contracts with our AI subprocessors — OpenAI, Microsoft, and Mistral — prohibit them from doing so either.

What happens to my data if I delete a document?

It's removed from active systems immediately, and from backups and logs within the retention window you've configured (1-180 days).

Questions?

If you need more detail for your own security or vendor review, contact us at [email protected].